Cloudflare Workers AI

Models at the edge, one fetch away from your Worker.

Workers AI runs inference at Cloudflare’s edge, so a Worker can call a model without a round trip to a regional GPU farm. The free unit is 10,000 neurons a day, shared across the catalogue. Llama 3.3 70B, GPT-OSS 120B and Qwen 3.8 stay on that tier.

The longest free window is 262K, on Qwen 3.8 27B — there is no 10M model here. Kimi K2.6, GLM 5.x and DeepSeek V4 are Workers Paid only. The REST URL needs your {account_id}; inside a Worker, use the AI binding instead.

Free models: 40
Max context: 262K
Free tier: 10,000 neurons per day
Requirement: Email account, no card

Endpoint

Base URLhttps://api.cloudflare.com/client/v4/accounts/{account_id}/ai/v1

Env varCLOUDFLARE_API_TOKEN

Rate limit10K neurons per day (shared across models)

OpenAI-compatible
text
image
video
code
reasoning

Code examples

npm install openai

import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/v1',
  apiKey: process.env.CLOUDFLARE_API_TOKEN
})

const response = await client.chat.completions.create({
  model: '@cf/openai/gpt-oss-120b',
  messages: [{ role: 'user', content: 'Explain closures in one paragraph.' }]
})

console.log(response.choices[0].message.content)

pip install openai

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/v1",
    api_key=os.environ["CLOUDFLARE_API_TOKEN"],
)

response = client.chat.completions.create(
    model="@cf/openai/gpt-oss-120b",
    messages=[{"role": "user", "content": "Explain closures in one paragraph."}],
)

print(response.choices[0].message.content)

curl https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  -d '{
    "model": "@cf/openai/gpt-oss-120b",
    "messages": [{"role": "user", "content": "Explain closures in one paragraph."}]
  }'

Worth knowing

Swap {account_id} in the base URL for your own account id.
Inside a Worker, use the AI binding instead — it skips the network hop and the token.
Kimi K2.6, GLM 5.2/5.3 and DeepSeek V4 are Workers Paid only. The longest free window is 262K on Qwen 3.8 27B.

Models you get for free39

GPT-OSS 120B@cf/openai/gpt-oss-120b
GPT-OSS 20B@cf/openai/gpt-oss-20b
Qwen 3.8 27B@cf/qwen/qwen3.8-27b
Llama 3.3 70B@cf/meta/llama-3.3-70b-instruct-fp8-fast
Llama 4 Scout@cf/meta/llama-4-scout-17b-16e-instruct
Llama 3.1 8B Fast@cf/meta/llama-3.1-8b-instruct-fast
Llama 3.1 8B FP8@cf/meta/llama-3.1-8b-instruct-fp8
Llama 3.2 11B Vision@cf/meta/llama-3.2-11b-vision-instruct
Llama 3.2 1B@cf/meta/llama-3.2-1b-instruct
Llama 3.2 3B@cf/meta/llama-3.2-3b-instruct
Gemma 4 26B@cf/google/gemma-4-26b-a4b-it
GLM 4.7 Flash@cf/zai-org/glm-4.7-flash
Granite 4.0 H Micro@cf/ibm/granite-4.0-h-micro
Nemotron 3 Super@cf/nvidia/nemotron-3-120b-a12b
Qwen 3 30B A3B@cf/qwen/qwen3-30b-a3b-fp8
Qwen 2.5 Coder 32B@cf/qwen/qwen2.5-coder-32b-instruct
QwQ 32B@cf/qwen/qwq-32b
Mistral Small 3.1@cf/mistralai/mistral-small-3.1-24b-instruct
DeepSeek R1 Distill 32B@cf/deepseek-ai/deepseek-r1-distill-qwen-32b
Llama Guard 3 8B@cf/meta/llama-guard-3-8b
Whisper Large V3 Turbo@cf/openai/whisper-large-v3-turbo
BGE M3@cf/baai/bge-m3
BGE Base@cf/baai/bge-base-en-v1.5
EmbeddingGemma 300M@cf/google/embeddinggemma-300m
FLUX.1 Schnell@cf/black-forest-labs/flux-1-schnell
Qwen 3 Embedding 0.6B@cf/qwen/qwen3-embedding-0.6b
Gemma 3 12B@cf/google/gemma-3-12b-it
Llama 3.1 8B@cf/meta/llama-3.1-8b-instruct
Mistral 7B@cf/mistralai/mistral-7b-instruct-v0.2
LLaVA 1.5 7B@cf/llava-hf/llava-1.5-7b-hf
Whisper@cf/openai/whisper
Nova 3@cf/deepgram/nova-3
MeloTTS@cf/myshell-ai/melotts
Stable Diffusion XL@cf/stabilityai/stable-diffusion-xl-base-1.0
Resnet 50@cf/microsoft/resnet-50
BART Large CNN@cf/facebook/bart-large-cnn
DistilBERT SST 2@cf/huggingface/distilbert-sst-2-int8
SQLCoder 7B@cf/defog/sqlcoder-7b-2
Phi 2@cf/microsoft/phi-2